I agree that any placer needs to be faster and more accurate than manual method. Also, for a production environment, the feeders are the problem, and there is no good solution: You need a few ten of those, and even at a couple of hundred each, the investment becomes big. I haven't had to need to do the math, but I suspect that the feeder cost fast goes so high, that it makes sense to buy a real commercial machine, with support, warranty and all that.
However, I don't think the feeders are an issue in a DIY or prototype lab. In that environment, you typically would be building one or two boards at a time. Throughput is not an issue. What the machine needs is a big enough work area and tape holders. I can see a system where you would have many slots where the machine can pick components along the whole slot. You would peel the cover tape off manually, load the slots and tell the machine that "C1-32 (100n, 0805) are in slot 1, positions 1-32", "R125-130 (2k2, 0603) are in slot 5, positions 5-10" and so on. Add to this the capability of picking up loose components (with help of down-looking camera and clever software) and an up-looking camera for BGAs. I think a simple DIY kit/prototype level machine is viable and has its place in the world. None exists yet, though.
But it still needs to be fast and accurate enough.